Eight methodological issues have led to the historical difficulties in dissecting the molecular genetics of MDD (a common, complex trait of enormous public health significance). GAIN inherently corrects two problems - inadequate statistical power and sparse, candidate-gene based genotyping. Additionally, first, most MDD samples are inhomogeneous in subject ancestry and are concatenated from disparate studies. Second, most MDD collections are highly selected samples of convenience containing unknown but potentially disastrous biases. Third, "controls" used in most association studies are highly dissimilar to cases in ascertainment and assessment and may not be at low likelihood for MDD. Fourth, minimal phenotypes are usually available - typically self-reported and exclude biological markers that can identify more homogeneous subgroups. Fifth, replication has historically been conducted external to a study and not within the study. Most critically, sixth, MDD cases are not usually directly evaluated by psychiatrists or clinical psychologists affiliated with the study - instead, non-psychiatrist research assistants gather the primary data with expert review occurring based on written records or only in reliability sub-studies. We propose to remediate all of these difficulties using existing samples from the Netherlands (1,860 cases with MDD and 1,860 matched controls). All subjects are participants in two large, longitudinal, population-based studies. The parent studies are closely coordinated, samples are from a known epidemiological sampling frame, controls are well-matched to cases, and multiple biomarkers are available for cases and controls (diurnal cortisol variation, thyroid function, and autonomic nervous system assessment, fMRI, and lymphocyte microarrays). This is arguably the best and most comprehensive sample collection for MDD in the world. [unreadable] [unreadable] [unreadable]